def testNoOp(self):
# No random cropping is performed since the size is value.shape.
for shape in (2, 1, 1), (2, 1, 3), (4, 5, 3):
value = np.arange(0, np.prod(shape), dtype=np.int32).reshape(shape)
with self.cached_session():
crop = random_ops.random_crop(value, shape).eval()
self.assertAllEqual(crop, value)
def testNoOp(self):
# No random cropping is performed since the size is value.shape.
for shape in (2, 1, 1), (2, 1, 3), (4, 5, 3):
value = np.arange(0, np.prod(shape), dtype=np.int32).reshape(shape)
with self.cached_session():
crop = random_ops.random_crop(value, shape).eval()
self.assertAllEqual(crop, value)
def image_augmentation(inputs, batch_size):
"""image augmentation."""
img = inputs
img = image_ops_impl.resize_images_v2(img, size=[224, 224])
img = random_ops.random_crop(img, size=[batch_size, 224, 224, 3])
img = rotate(img)
img = zoom(img)
return img
def testContains(self):
with self.cached_session():
shape = (3, 5, 7)
target = (2, 3, 4)
value = np.random.randint(1000000, size=shape)
value_set = set(
tuple(value[i:i + 2, j:j + 3, k:k + 4].ravel())
for i in range(2) for j in range(3) for k in range(4))
crop = random_ops.random_crop(value, size=target)
for _ in range(20):
y = self.evaluate(crop)
self.assertAllEqual(y.shape, target)
self.assertTrue(tuple(y.ravel()) in value_set)
def testContains(self):
with self.cached_session():
shape = (3, 5, 7)
target = (2, 3, 4)
value = np.random.randint(1000000, size=shape)
value_set = set(
tuple(value[i:i + 2, j:j + 3, k:k + 4].ravel())
for i in range(2) for j in range(3) for k in range(4))
crop = random_ops.random_crop(value, size=target)
for _ in range(20):
y = self.evaluate(crop)
self.assertAllEqual(y.shape, target)
self.assertTrue(tuple(y.ravel()) in value_set)
def random_crop(labeled_tensor, shape_map, seed=None, name=None):
"""Randomly crops a tensor to a given size.
See tf.random_crop.
Args:
labeled_tensor: The input tensor.
shape_map: A dictionary mapping axis names to the size of the random crop
for that dimension.
seed: An optional random seed.
name: An optional op name.
Returns:
A tensor of the same rank as `labeled_tensor`, cropped randomly in the
selected dimensions.
Raises:
ValueError: If the shape map contains an axis name not in the input tensor.
"""
with ops.name_scope(name, 'lt_random_crop', [labeled_tensor]) as scope:
labeled_tensor = core.convert_to_labeled_tensor(labeled_tensor)
for axis_name in shape_map:
if axis_name not in labeled_tensor.axes:
raise ValueError('Selection axis %s not in axes %s' %
(axis_name, labeled_tensor.axes))
shape = []
axes = []
for axis in labeled_tensor.axes.values():
if axis.name in shape_map:
size = shape_map[axis.name]
shape.append(size)
# We lose labels for the axes we crop, leaving just the size.
axes.append((axis.name, size))
else:
shape.append(len(axis))
axes.append(axis)
crop_op = random_ops.random_crop(labeled_tensor.tensor,
shape,
seed=seed,
name=scope)
return core.LabeledTensor(crop_op, axes)
def random_crop(labeled_tensor, shape_map, seed=None, name=None):
"""Randomly crops a tensor to a given size.
See tf.random_crop.
Args:
labeled_tensor: The input tensor.
shape_map: A dictionary mapping axis names to the size of the random crop
for that dimension.
seed: An optional random seed.
name: An optional op name.
Returns:
A tensor of the same rank as `labeled_tensor`, cropped randomly in the
selected dimensions.
Raises:
ValueError: If the shape map contains an axis name not in the input tensor.
"""
with ops.name_scope(name, 'lt_random_crop', [labeled_tensor]) as scope:
labeled_tensor = core.convert_to_labeled_tensor(labeled_tensor)
for axis_name in shape_map:
if axis_name not in labeled_tensor.axes:
raise ValueError('Selection axis %s not in axes %s' %
(axis_name, labeled_tensor.axes))
shape = []
axes = []
for axis in labeled_tensor.axes.values():
if axis.name in shape_map:
size = shape_map[axis.name]
shape.append(size)
# We lose labels for the axes we crop, leaving just the size.
axes.append((axis.name, size))
else:
shape.append(len(axis))
axes.append(axis)
crop_op = random_ops.random_crop(labeled_tensor.tensor,
shape,
seed=seed,
name=scope)
return core.LabeledTensor(crop_op, axes)
def testRandomization(self):
# Run 1x1 crop num_samples times in an image and ensure that one finds each
# pixel 1/size of the time.
num_samples = 1000
shape = [5, 4, 1]
size = np.prod(shape)
single = [1, 1, 1]
value = np.arange(size).reshape(shape)
with self.cached_session():
crop = random_ops.random_crop(value, single, seed=7)
counts = np.zeros(size, dtype=np.int32)
for _ in range(num_samples):
y = self.evaluate(crop)
self.assertAllEqual(y.shape, single)
counts[y] += 1
# Calculate the mean and 4 * standard deviation.
mean = np.repeat(num_samples / size, size)
four_stddev = 4.0 * np.sqrt(mean)
# Ensure that each entry is observed in 1/size of the samples
# within 4 standard deviations.
self.assertAllClose(counts, mean, atol=four_stddev)
def testRandomization(self):
# Run 1x1 crop num_samples times in an image and ensure that one finds each
# pixel 1/size of the time.
num_samples = 1000
shape = [5, 4, 1]
size = np.prod(shape)
single = [1, 1, 1]
value = np.arange(size).reshape(shape)
with self.cached_session():
crop = random_ops.random_crop(value, single, seed=7)
counts = np.zeros(size, dtype=np.int32)
for _ in range(num_samples):
y = self.evaluate(crop)
self.assertAllEqual(y.shape, single)
counts[y] += 1
# Calculate the mean and 4 * standard deviation.
mean = np.repeat(num_samples / size, size)
four_stddev = 4.0 * np.sqrt(mean)
# Ensure that each entry is observed in 1/size of the samples
# within 4 standard deviations.
self.assertAllClose(counts, mean, atol=four_stddev)
